DESIGN AND DEVELOPMENT OF HUMAN COMPUTER INTERFACE USING ELECTROOCULOGRAM WITH DEEP LEARNING.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Human-Computer Interaction

Background:

• Life assistive devices are crucial for communication.
• Electrooculogram (EOG) based Human-Computer Interfaces (HCI) offer improved performance and accuracy over conventional methods.
• Developing robust EOG-based HCI requires effective signal processing and feature extraction techniques.

Purpose of the Study:

• To design and evaluate a nine-state EOG-based HCI system.
• To compare the efficacy of band power and Hilbert Huang Transform (HHT) for feature extraction.
• To determine the optimal method for classifying and recognizing single-trial eye movements for HCI applications.

Main Methods:

• Acquired EOG signals from twenty subjects using a five-electrode system to capture horizontal and vertical eye movements.
• Preprocessed EOG signals to remove artifacts and extracted features using band power and HHT.
• Trained a Pattern Recognition Neural Network (PRNN) to classify tasks based on extracted features.

Main Results:

• The PRNN achieved classification accuracies of 92.17% for band power and 91.85% for HHT.
• Band power feature extraction with PRNN demonstrated superior classification and recognition accuracy compared to HHT.
• Male subjects exhibited higher performance and accuracy than female subjects, with the 26-32 age group showing the best results.

Conclusions:

• Band power combined with PRNN is an effective method for developing accurate EOG-based HCI systems.
• The developed system shows potential for practical application in life assistive devices.
• Performance variations were observed across genders and age groups, suggesting areas for future optimization.